Some questions about CWRES-IVAR and CWRES-TAD

Phoenix Certara Forums Modelling and Simulation
1

Dear all，

Thanks for your reading！I’m a student who is learning population pharmacokinetic modeling using Phoenix NLME. Now I have trouble in the model diagnostic diagram CWRES-IVAR and CWRES-TAD. In my diagrams the blue line always has a trend but the other model evaluation methods can be able to meet standards, and there’s nothing I can do to improve it. Can you give me some advice?

Hope for your answers! Grateful!

2

It’s hard to say without seeing your data and any models you have tried. Firstly I would look at whether these trends can be managed by a different residual error models e.g. try Multiplicative, or Poisson if your concentrations range over several orders of magnitude.

I don’t know if you have considered covariates yet but sometimes addingthose in starts to explain better the residual variability.

Also, if you are modelling multiple doses and/or different dose levels then you may be seeing non-linear kinetics

Simon.
PS you may want to look at https://www.certarauniversity.com/store/617231-101-od-introduction-to-population-modeling-using-phoenix-nlme

or the bundle; https://www.certarauniversity.com/store/3400-b123-lo-beginner-population-modeling-bundle-with-2-exams?is_bundle=1

3

Appreciate your reply.

EDIT by Simon. Hi XQLI, what browser are you using? Try edit the post and then choose full editor;

Browse.jpg
Browse.jpg1014×756 143 KB

4

This tutorial paper; CPT Pharmacometrics Syst. Pharmacol. (2017) 6, 87–109, may be useful too.

Model Evaluation of Continuous Data PharmacometricModels: Metrics and Graphics

THT Nguyen1, M-S Mouksassi2, N Holford3, N Al-Huniti4, I Freedman5, AC Hooker6, J John7, MO Karlsson6, DR Mould8,JJ Perez Ruixo9, EL Plan10, R Savic11, JGC van Hasselt12, B Weber13, C Zhou14, E Comets1,15and F Mentre1*for the Model Evaluation Group of the International Society of Pharmacometrics (ISoP) Best Practice Committee.

https://ascpt.onlinelibrary.wiley.com/doi/pdf/10.1002/psp4.12161

Simon.

5

Thank you for your reply.It is very useful to me.

6

I’m sorry for replying to you so late. These are some of my results. I don’t know how to improve my model and I’m looking forward to your suggestions.

结果图.pdf (70.1 KB)

7

XQLI, an IVAR plot that shows this type of deviation from the central line suggests an issue with your structural model. I suggest you look at that again and consider an extra compartment or perhaps a tlag. Simon.

ivar_structural.jpg
ivar_structural.jpg620×644 76.8 KB

8

Thank you for your advice. It’s useful to me. I’ll give it a try.

9

[quote=“Simon Davis, username:sdavis”]

XQLI, an IVAR plot that shows this type of deviation from the central line suggests an issue with your structural model. I suggest you look at that again and consider an extra compartment or perhaps a tlag. Simon.

[/quote]I’m sorry to bother you again. After modifying the structural model, the CWRES correlation graph looks like this. Do I need to continue to modify the structural model? Or is it time to start looking for covariates to build models of covariates?

I am sorry that I am a self-learner and there is no one around to help me, so I can only ask for your help. Looking forward to your reply.
模型结果.pdf (51.9 KB)

10

that looks pretty good to me. I would start looking if ny covariates can be justified, Simon.

11

Thank you very much for your reply Simon. Best wishes to you.

12

Excuse me again Simon, I venture to ask, after the addition of covariator model and covariance model, there is no significant improvement in the CWRES related image, can the current image be used for article publication? XQLI

13

You cannot judge a model on one diagnostic alone, you need to consider -2LL,AIC, use tools lik bootstrap and VPC as well. That said I think your CWRES plot looks like the best of the different models you have shared so far, in the end the choice will be yours

SImon

14

I get it. Thank you so much.

XQLI